Abstract:
A computer-implemented method for restoring a sequence for a dataset with frame dropping includes receiving an input sequence. A set of features is extracted from the input sequence. A frequency distribution is determined for the input sequence based on the extracted features. Time domain information for the sequence is restored and in turn, data for the input sequence is augmented based on the restored time domain information. Additionally, noise is removed from the input sequence.
Abstract:
Certain aspects of the present disclosure provide techniques for improved machine learning using gradient pruning, comprising computing, using a first batch of training data, a first gradient tensor comprising a gradient for each parameter of a parameter tensor for a machine learning model; identifying a first subset of gradients in the first gradient tensor based on a first gradient criteria; and updating a first subset of parameters in the parameter tensor based on the first subset of gradients in the first gradient tensor.
Abstract:
Techniques for access terminal radio link monitoring on a shared communication medium are disclosed. In an aspect, an access terminal detects a missed reference signal event associated with a radio link established on the shared communication medium, wherein detecting the missed reference signal event comprises determining that the access terminal did not detect a reference signal for measuring a quality of the radio link transmitted during a reference signal configuration window, assigns an error metric to the missed reference signal event based on reference signal monitoring capabilities of the access terminal, and triggers a radio link failure based on the assigned error metric. In an aspect, the missed reference signal event may be a missed Discovery Reference Signaling (DRS) event, the error metric may be a Block Error Rate (BLER) weight, and the reference signal for measuring the quality of the radio link comprises a Cell-specific Reference Signal (CRS).
Abstract:
The present disclosure relates generally to uplink procedures on a shared communication medium. In an aspect, an access terminal receives a downlink subframe from an access point on the shared communication medium and, in response to receiving the downlink subframe, transmits uplink control information (UCI) for the downlink subframe on a first uplink subframe of a first UCI channel of a plurality of UCI channels.
Abstract:
Techniques for managing preamble transmission and processing on a shared communication medium are disclosed. An access point or an access terminal, for example, may generate a preamble for silencing communication on a communication medium with respect to an upcoming data transmission, configure the preamble to identify one or more target devices for the silencing, and transmit the preamble over the communication medium in advance of the data transmission. Conversely, the access point or the access terminal may receive a preamble (as a receiving device) over a communication medium, identify one or more target devices for silencing communication on the communication medium with respect to an upcoming data transmission based on the preamble, and selectively silence communication over the communication medium based on itself (as the receiving device) being among the one or more target devices.
Abstract:
Techniques for managing access to a shared communication medium are disclosed. Scheduling grants may be sent to different access terminals for different sets of resources for uplink transmission on the communication medium. A series of re-contention gaps may be scheduled for access terminal contention within or between the different sets of resources. Uplink and downlink transmission on the communication medium may be silenced during each of the series of re-contention gaps. Moreover, an access terminal may receive a scheduling grant that allocates a set of resources to the access terminal for uplink transmission on a communication medium and contend for access to the communication medium based on the scheduling grant. The access terminal may then selectively transmit uplink traffic over the allocated set of resources based on the contending.
Abstract:
A method for enabling an active hand-in from a macro base station network to a femtocell network includes servicing an active hand-in of a mobile entity from a macro base station to a femtocell network, using a first femtocell of the femtocell network. The active hand-in includes a hard handoff of the mobile entity from the macro base station with soft handoff of the mobile entity enabled between the first femtocell and one or more neighboring femtocells in the femtocell network. The hard handoff with soft handoff enabled may be implemented using novel procedures implemented by one or more entities of a wireless communications network including the femtocells and macro base station.
Abstract:
The present disclosure presents a method and apparatus for joint power and resource management in a wireless network. For example, the disclosure presents a method for receiving reference signal received power (RSRP) measurements of one or more neighboring base stations of a base station. In addition, such an example method, may include calibrating a transmit power of the base station based at least on the received measurements, and adjusting transmit resources of the base station in response to the calibration. As such, joint power and resource management in a wireless network may be achieved.
Abstract:
Certain aspects provide techniques and apparatuses for efficiently processing inputs in a neural network using multiple receptive field sizes. An example method includes partitioning a first input into a first set of channels and a second set of channels. At a first layer of a neural network, the first set of channels and the second set of channels are convolved into a first output having a smaller dimensionality a dimensionality of the first input. The first set of channels and the first output are concatenated into a second input. The second input is convolved into a second output via a second layer of the neural network, wherein the second output merges a first receptive field generated by the first layer with a larger second receptive field generated by the second layer. One or more actions are taken based on at least one of the first output and the second output.
Abstract:
A processor-implemented method for adaptive quantization in an artificial neural network (ANN) includes receiving an ANN model. The ANN model has multiple channels of target activations. A quantization module is incorporated between a first linear layer of the ANN and a second linear layer of the ANN to generate an adapted ANN. The quantization module scales a first set of weights and biases of the first linear layer based on a learnable quantization module parameter and scales a second set of weights of the second linear layer based on an inverse of the learnable quantization module parameter.